The Essence of Moats for AI Startups: Seven Powers and Practical Strategy

Brief Summary: This video examines what AI startups should establish as their "moat" to survive competitively, and when and how to think about moats during the startup journey. The core framework is Hamilton Helmer's "Seven Powers," illustrated with real AI startup examples showing how each power applies. The ultimate conclusion: the most powerful moat is "speed," and the other powers should be considered strategically only after a certain stage.

1. What Is a Moat, and Why Has It Become More Important?

The video opens by discussing how the concept of a "moat" has become increasingly common among aspiring founders in the AI era. A "moat" originally refers to a defensive trench, but here it means "a barrier to entry or competitive advantage that competitors can't easily replicate."

"A moat is fundamentally a defensive concept. If you have nothing to defend, there's nothing to worry about in the first place."

After the meme spread that AI (especially ChatGPT) is "easy to copy," the anxiety that "AI agent companies will just get caught up eventually" became widespread.

But the hosts emphasize that "AI startups can absolutely build deep and compelling moats," and proceed to examine real cases.

"There's one particularly important book: Hamilton Helmer's Seven Powers."

2. Hamilton Helmer's 'Seven Powers' and the 7 Moat Categories

Published in 2016, the book primarily uses examples from 2000s-era internet companies (Oracle, Facebook, Netflix, etc.). The video applies these "7 powers" freshly to today's AI startup reality.

"Essentially, it covers seven moat categories. The strategic advantages a business can have converge to a few types. The essence stays the same even as times change."

The point is also made that "the more intense market competition becomes, the more margins converge to zero without a moat — making this a question of business survival itself."

3. When and How to Think About Moats (Stage-by-Stage Startup Strategy)

The advice is to be careful not to get paralyzed by thinking about moats too early.

"In the beginning, find a real customer pain point and solve it. That's the first step of entrepreneurship."

In most cases, "speed" is the only moat, and "getting a product that actually works to market fast" is what matters most at a certain stage.

"Executing faster than anyone is what differentiates you from big companies. Startups can ship a feature in a day without meetings or bureaucracy."

For example, a company called Cursor shipped features daily (edit: daily sprints), handling in moments what would take big companies weeks to years.

After establishing that "speed is the first moat," the emphasis is that the other moats should be considered step by step after initial growth (i.e., after market validation).

"If you have nothing worth defending yet and you're worrying about moats, that's like stressing over a puddle in a field. Wait until you've built something worth attacking."

4. The 7 Powers and Real AI Startup Examples

1) Process Power

"Complex operations, technology, and workflows that nobody can easily imitate" become the moat.
Examples: Plaid (fintech integration), Greenlight/Kasa (banking KYC, loan AI) — cases requiring far more reliability than a hackathon demo, with complex and repetitive processes.

"You can't replicate this with a 7-day hackathon demo. You need to build massive infrastructure — servers, data, edge case support — brick by brick."

2) Cornered Resources

Resources with specific barriers to entry — regulation, patents, networks — become the moat.
Examples: Government and defense (DoD, Palantir, Scale AI) contracts requiring high-level connections and certifications, or proprietary data from customized data and workflows.

"Cornered resources aren't just diamond mines — they can include the 'brain space' of our customers."

In AI, specialized proprietary models (high-performance LLMs specialized for specific tasks) can be the ultimate cornered resource.

3) Switching Costs

The force that makes it hard for customers to move to another solution once they're established.
Examples: Oracle databases, Salesforce — B2B services where data and workflow customization make "switching" extremely difficult.
In AI, customized ML workflows and company-specific customization are representative.

"Switching costs are still important in the AI era. LLMs and codegen can make data migration easier. But the more customized processes accumulate — and the more usage history builds up — the higher the switching costs."

"Your own prompt sets, data-based personalization, persistent memory — these all become switching costs even for consumers."

4) Counter Positioning

Adopting a model or strategy that established players absolutely cannot follow (because following it would destroy their existing business).
Example: When AI reduces headcount, existing SaaS companies (with per-seat pricing) take a hit from enterprise customers. AI-native startups compete with entirely different pricing models or feature specializations.

"Counter positioning is an extremely powerful strategic weapon when incumbents would have to cannibalize their own business."

"OpenAI's ChatGPT growing frighteningly fast through 'speed' and 'simplicity,' forcing Google into a defensive posture, is a classic intuitive example."

Sometimes the latecomer surpasses #1 through faster learning and productization (Stripe, DoorDash, Lorra vs Harvey, etc.)
When AI proves superior to humans in a given area, an entirely new market structure opens up

5. Network Effects & Scale Economies: New Moats for the AI Era

5) Network Effects

As users and customers grow, the value of the product/service increases exponentially.
Traditional examples: Facebook, Visa cards, etc.
AI version: User data and feedback directly improve the model → continuously upgrading service = network effects!

"Cursor — as users grow, more code usage data accumulates, which feeds back into autocomplete quality, creating a structure where the winner keeps pulling ahead."

"In-house custom workflows, learned prompts, iterative improvement through 'evals' — these can only be created as real users grow."

6) Scale Economies

Building large-scale market presence/infrastructure to gain the structural advantage of "lower unit costs."
Traditional examples: Large logistics companies (UPS, Amazon)
In AI, this is especially prominent at the "model layer" (mega-scale LLMs, etc.)
Example: DeepSeek released innovations that dramatically lowered LLM training costs, potentially shaking the fundamentals of scale economies

"Previously, training mega-scale models required enormous capital and infrastructure, but a single innovative new approach could shake the entire market structure."

Application examples: Exa, Channel3, Orange Slice, etc., rapidly building web crawling infrastructure for search (→ Once a large-scale crawl is done, it can be infinitely extended into similar services)

6. Again, the Most Powerful Moat: "Speed" and Solving Real "Pain"

Finally, the core message of the moat discussion is "First, you need to be able to quickly solve someone's truly urgent problem (real pain, practical pain point) — and that is a startup's absolute advantage and the ultimate moat."

"When you solve a problem that's literally a matter of survival for customers, that's the moat that creates 0→1. Don't limit yourself prematurely with moat frameworks."

Conclusion

For AI startups, a moat means "speed" and "relentless problem-solving" as the absolute advantage — the first true defensive wall. Once you've established yourself in the market, apply Hamilton Helmer's Seven Powers framework (process, cornered resources, switching costs, etc.) to your specific business structure. But above all, never forget that the most important thing is "solving real problems faster than everyone else"!

"The first thing you need to do is find someone in pain and solve their problem fast. Everything else can be figured out later!"

May your AI startup journey be accompanied by a small but sturdy moat.